A multi-threaded processor that takes one or even multiple tile requests and can process them iteratively. More...

#include <mlMultiThreadedPageRequestProcessor.h>

Inheritance diagram for ml::MultiThreadedPageRequestProcessor:

Public Member Functions
	MultiThreadedPageRequestProcessor (Host *host=nullptr)

	~MultiThreadedPageRequestProcessor () override

bool	supportsMultiThreading () override
	Returns if multi-threading is supported.

void	setNumWorkerThreads (std::size_t threads) override
	Set number of worker threads.

void	addRootTileRequest (TileRequest *tileRequest) override
	Adds the root `tileRequest` that should be processed (the ownership stays with the caller).

void	processAll () override
	Processes all requests until the cursors have traversed the whole tree and the queue is empty.

void	process (double timeBudget) override
	Process the requests for the given timeBudget given in seconds.

void	launchMultiThreading ()
	Launch the multi-threading.

bool	needsProcessing () override
	Returns if the tile request needs some more processing.

void	append (PageRequest *request) override
	Implements PageRequestQueueInterface interface.

void	enterProcessingScope () override
	Enters a recursive processing scope (always called from main thread!)

void	leaveProcessingScope () override
	Leaves a recursive processing scope (always called from main thread!)

bool	processingWasSingleThreaded () const override
	Processing is always single-threaded, only implemented for testing purposes.

void	setAllowLateMultiThreadingStart (bool flag)
	This can be set to false to debug single threaded part of this class...

Public Member Functions inherited from ml::PageRequestProcessorBase
	PageRequestProcessorBase (Host *host)

	~PageRequestProcessorBase () override

void	removeCancelledRequests () override
	Removes all requests that are canceled and no longer needed.

ProcessingTimeLine &	getProcessingTimeLine ()
	Get access to the processing time line.

ProcessingTimeLine::TimeLine *	getOverviewTimeLine ()
	Get access to the overview time line (only if time line profiling is on)

void	enableTimeLineProfiling (bool flag)
	Enable/disable time line profiling.

bool	isTimeLineProfilingEnabled () const
	Returns if time line profiling is enabled.

int	getProcessingScope () const
	Return the current processing scope.

void	enterProcessingScope () override
	Enters a recursive processing scope (always called from main thread!)

void	leaveProcessingScope () override
	Leaves a recursive processing scope (always called from main thread!)

Public Member Functions inherited from ml::PageRequestProcessor
	~PageRequestProcessor () override
	Virtual destructor.

Public Member Functions inherited from ml::PageRequestQueueInterface
virtual	~PageRequestQueueInterface ()
	Virtual destructor to get rid of warning.

Additional Inherited Members
Static Public Member Functions inherited from ml::PageRequestProcessor
static void	enableTileRequestAllocationFailure (int count=1)
	Enables failing on every `count` allocation.

static void	disableTileRequestAllocationFailure ()
	Disables allocation failure (default!)

Protected Member Functions inherited from ml::PageRequestProcessorBase
bool	createMoreWork (bool sleepWhenBlocked, bool allBlocked=nullptr, MLint pageRequestLimit=100000, MLint resultPageRequestCount=nullptr)
	Creates more work items, returns true if new work was created.

void	setupCurrentProcessingScope ()
	helper

void	terminateAllRequestsInGUIThread ()
	terminate all requests

void	addNewCursor (PageRequestCursor *cursor)
	add the new cursor

UserThreadData *	getUserThreadData (PageRequest *request, PerThreadStorage &perThreadStorage)
	Gets or allocates the UserThreadData for the given request as part of the perThreadStorage.

Protected Attributes inherited from ml::PageRequestProcessorBase
Host *	_host
	optional pointer to the host

bool	_terminated
	flag if the calculation was terminated

int	_rootTileRequests
	the number of root tile requests, if this gets down to 0, we are done

std::vector< std::vector< PageRequestCursor * > >	_cursors
	The cursors that are handled by the producer thread, one list for each processing scope.

int	_processingScope
	the current processing scope (0 = BG tasks, 1 = normal get tile, >1 = recursive get tile

int	_producerProcessingScope
	the current cursor scope of the producer

int	_producerCursorCounter
	how many cursors the producer contains in all lists

int	_guiThreadRecursionCounter
	the recursion of the GUI thread (0 = outer entry)

ScopeStackEntry *	_currentScope
	the current scope pointer

std::vector< ScopeStackEntry >	_scopeStack
	Stack of processing scopes for performance reasons, we only grow this stack, we never shrink it.

TimeCounter	_progressUpdateTimer
	Timer used for progress update from GUI thread.

PerThreadStorage	_guiPerThreadStorage
	The per thread storage for the GUI thread.

bool	_useTimeLineProfiling

ProcessingTimeLine	_profile

ProcessingTimeLine::TimeLine *	_guiThreadTimeLine

ProcessingTimeLine::TimeLine *	_overviewTimeLine

Detailed Description

A multi-threaded processor that takes one or even multiple tile requests and can process them iteratively.

Definition at line 45 of file mlMultiThreadedPageRequestProcessor.h.

Constructor & Destructor Documentation

◆ MultiThreadedPageRequestProcessor()

ml::MultiThreadedPageRequestProcessor::MultiThreadedPageRequestProcessor ( Host * host = nullptr )

◆ ~MultiThreadedPageRequestProcessor()

ml::MultiThreadedPageRequestProcessor::~MultiThreadedPageRequestProcessor ( )

override

Member Function Documentation

◆ addRootTileRequest()

void ml::MultiThreadedPageRequestProcessor::addRootTileRequest ( TileRequest * tileRequest )

overridevirtual

Adds the root tileRequest that should be processed (the ownership stays with the caller).

Implements ml::PageRequestProcessor.

◆ append()

void ml::MultiThreadedPageRequestProcessor::append ( PageRequest * request )

overridevirtual

Implements PageRequestQueueInterface interface.

Implements ml::PageRequestQueueInterface.

◆ enterProcessingScope()

void ml::MultiThreadedPageRequestProcessor::enterProcessingScope ( )

overridevirtual

Enters a recursive processing scope (always called from main thread!)

Implements ml::PageRequestProcessor.

◆ launchMultiThreading()

void ml::MultiThreadedPageRequestProcessor::launchMultiThreading ( )

Launch the multi-threading.

◆ leaveProcessingScope()

void ml::MultiThreadedPageRequestProcessor::leaveProcessingScope ( )

overridevirtual

Leaves a recursive processing scope (always called from main thread!)

Implements ml::PageRequestProcessor.

◆ needsProcessing()

bool ml::MultiThreadedPageRequestProcessor::needsProcessing ( )

overridevirtual

Returns if the tile request needs some more processing.

Implements ml::PageRequestProcessor.

◆ process()

void ml::MultiThreadedPageRequestProcessor::process ( double timeBudget )

overridevirtual

Process the requests for the given timeBudget given in seconds.

Implements ml::PageRequestProcessor.

◆ processAll()

void ml::MultiThreadedPageRequestProcessor::processAll ( )

overridevirtual

Processes all requests until the cursors have traversed the whole tree and the queue is empty.

Implements ml::PageRequestProcessor.

◆ processingWasSingleThreaded()

bool ml::MultiThreadedPageRequestProcessor::processingWasSingleThreaded ( ) const

inlineoverridevirtual

Processing is always single-threaded, only implemented for testing purposes.

Implements ml::PageRequestProcessor.

Definition at line 81 of file mlMultiThreadedPageRequestProcessor.h.

◆ setAllowLateMultiThreadingStart()

void ml::MultiThreadedPageRequestProcessor::setAllowLateMultiThreadingStart ( bool flag )

inline

This can be set to false to debug single threaded part of this class...

Definition at line 84 of file mlMultiThreadedPageRequestProcessor.h.

References mlrange_cast().

◆ setNumWorkerThreads()

void ml::MultiThreadedPageRequestProcessor::setNumWorkerThreads ( std::size_t threads )

overridevirtual

Set number of worker threads.

Implements ml::PageRequestProcessor.

◆ supportsMultiThreading()

bool ml::MultiThreadedPageRequestProcessor::supportsMultiThreading ( )

inlineoverridevirtual

Returns if multi-threading is supported.

Implements ml::PageRequestProcessorBase.

Definition at line 52 of file mlMultiThreadedPageRequestProcessor.h.

The documentation for this class was generated from the following file:

MeVis/Foundation/Sources/ML/include/host/mlMultiThreadedPageRequestProcessor.h

Public Member Functions

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ MultiThreadedPageRequestProcessor()

◆ ~MultiThreadedPageRequestProcessor()

Member Function Documentation

◆ addRootTileRequest()

◆ append()

◆ enterProcessingScope()

◆ launchMultiThreading()

◆ leaveProcessingScope()

◆ needsProcessing()

◆ process()

◆ processAll()

◆ processingWasSingleThreaded()

◆ setAllowLateMultiThreadingStart()

◆ setNumWorkerThreads()

◆ supportsMultiThreading()